Agent-Based Model (ABM) — Core Framework

With no closed-form solution, the system state emerges from the interaction of hundreds or thousands of heterogeneous agents. Below are the agents' behavior rules and the market-clearing mechanism.

Agent Behavior Rule

\[ E_i[p_{t+1}] = f_i(p_t, p_{t-1}, F, \sigma_t, \ldots),\quad D_{i,t} = \beta_i \cdot \frac{E_i[p_{t+1}] - p_t}{\sigma_t^2} \]

Agent i forms an expected price based on its strategy f_i and, combined with risk preference β_i, determines demand D_{i,t}.

Price-Clearing Mechanism

\[ p_{t+1} = p_t + \frac{\lambda}{N}\sum_{i=1}^{N} D_{i,t} + \varepsilon_t,\quad \varepsilon_t \sim \mathcal{N}(0, \sigma_\varepsilon^2) \]

Aggregating the net demand of all agents, price is updated via the price adjustment speed λ, with ε_t as a random shock.

Wealth Dynamics Evolution

\[ w_{i,t+1} = w_{i,t} + D_{i,t} \cdot (p_{t+1} - p_t),\quad S_t = \{p_t, \{w_{i,t}\}_{i=1}^N, \sigma_t\} \]

Wealth updates with price changes, and the system state S_t is characterized by price, wealth distribution, and volatility.

💡 The core idea of ABM: Instead of relying on top-down analytical equations, it simulates the micro-behavior of each heterogeneous agent from the bottom up. Macro regularities (such as price fluctuations, wealth divergence, and market bubbles) are the result of local interactions among a large number of agents that emerge—precisely the complex dynamics that traditional mathematical models struggle to capture.
N: Number of Agents F: Fundamental Value λ: Price Adjustment Speed β_i: Risk Preference f_i(·): Strategy Function ε_t: Random Shock σ_t: Market Volatility w_{i,t}: Agent Wealth
Interactive Simulation — Financial Market ABM Demo

Adjust the left-side parameters and observe the emergent changes in price trends and wealth distribution in the right-side charts. Supports auto-run, single-step debugging, and reset.

30300
50200
0.051.00
0%60%
0%60%
0%60%
125
100.00
Current Price
0
Simulation Steps
0.00%
Volatility
0.000
Gini Coefficient
Market Price Fundamental Value F Price Fluctuation Range
Fundamentalists Trend Followers Noise Traders Contrarians
Deep Dive into ABM (Agent-Based Model)

From plain intuition to academic depth, gain a comprehensive understanding of ABM's value and limitations in economics and finance.

Plain Introduction

Imagine a simulated city: each "agent" is a trader with its own ideas—some firmly believe prices will revert to value, some chase rallies and sell on dips, and others act purely on intuition. They buy and sell among themselves in the market, and prices arise naturally like the outcome of a vote. ABM is such a digital sandbox, letting us observe in a computer how "crowd behavior" creates the booms, crashes, and wealth divergence commonly seen in real markets.

What Problems It Solves

Traditional economic models usually assume all participants are fully rational and homogeneous, ignoring the core feature of real markets—diversity. ABM allows each agent to have different strategies, information, and risk preferences, enabling it to:

  • Reproduce "fat-tailed" distributions (extreme events occur more often than predicted by the normal distribution)
  • Simulate the spread mechanisms of irrational exuberance and panic selling
  • Study the endogenous evolution of wealth inequality

Why It Matters

ABM is a key tool for understanding complex adaptive systems. In financial regulation, it is used for stress testing and systemic risk assessment; in macroeconomics, it helps researchers explore "how micro-motives lead to macro-outcomes". The 2022 Nobel Prize in Economics was awarded to Diamond & Dybvig for their bank-run model, and ABM is the natural extension that scales and dynamizes such insights.

Application Scenarios

  • Financial Markets:simulate high-frequency trading, circuit breakers, and liquidity crises
  • Macroeconomy:study monetary policy transmission and business-cycle emergence
  • Real Estate:simulate homebuyer behavior and housing-bubble formation
  • Supply Chain:analyze inter-firm risk contagion and resilience
  • Climate Economy:evaluate the impact of carbon-tax policy on firms and consumers

Limitations

  • Calibration difficulty:numerous model parameters, hard to estimate precisely from real data
  • Computational cost:large-scale simulation requires substantial computing resources
  • Interpretability:the causes of emergent results are sometimes hard to trace
  • Overfitting risk:may "manufacture" desired results rather than real patterns
  • Lack of unified framework:different ABM models are hard to compare directly

Standard Introduction

The Agent-Based Model (ABM) is a computational modeling method that simulates the dynamic evolution of a large number of agents in a computer by defining the behavior rules f_i(·) and interaction protocols of heterogeneous agents. Unlike traditional equilibrium analysis, ABM does not assume the system reaches equilibrium; instead, it lets macro phenomena (such as price series, wealth distribution, and market volatility) emerge naturally from the local interactions of micro-agents. This method is widely used in economics, sociology, ecology, and epidemiology.